Retail AI ambition is 8.2/10. Readiness is 5.1/10. That’s the whole story.

The Validify × WPP Enterprise Solutions 2026 Retail AI Readiness Report landed this month and it’s the clearest data yet on a pattern we’ve been seeing in every retail engagement for the last two years. Retailers want AI. Their organisations aren’t ready to deploy it. The gap between those two facts is where most retail AI investment quietly evaporates.

The report isn’t a doom loop. It’s an operating manual for closing the gap. Here’s the version through our lens.

The numbers that matter

Validify’s retail community was surveyed on their AI maturity. Openness to AI: 8.2 out of 10 — very high. LLM familiarity: 5.5. AI understanding: 5.9. Data readiness: 5.1. Confidence in ROI: 5.0. Every foundational capability sits materially below the ambition.

The strategy picture is worse. Only 17.5% of respondents have a defined AI strategy or governance framework. 36.5% are developing one. 38.1% have no formal strategy but are “exploring opportunities” — which usually means fragmented pilots. 52% expect their AI budget to increase in 2026 anyway.

WPP’s wider transformation research adds context: 64% of transformation projects start without a clear roadmap. 59% of respondents believe their organisation’s data practices are not mature enough to support advanced digital technologies. 58% lack the governance frameworks to accelerate safely. 68% say legacy systems hinder transformation.

If you’re a retail CDO reading that sequence and thinking “that’s us” — you’re not the exception. You’re the pattern.

The readiness gap is a data foundation gap

The report’s Section 4 makes an observation that shouldn’t be surprising but constantly is: the retail AI use cases attracting the most investment (AI-native search, discovery, conversion — 67% of respondents rank it top priority) depend on foundations that sit lower in the priority list (data readiness and architecture — 44%). The visible layer is prioritised. The layer that makes the visible layer work isn’t.

This is what stalls the projects. An AI initiative that has to draw from a data warehouse where nobody knows the source of truth, where PII sits in tables it shouldn’t, where 58% of stored procedures are orphaned (a number we found in a recent Snowflake audit for a UK retailer), and where dev warehouses run 24×7 for no defensible reason — that initiative doesn’t fail at the model. It fails at the substrate.

We call this waste, risk and rot. Waste is over-provisioned compute and forgotten tables paying storage for nothing. Risk is unreviewed roles, absent MFA, PII in the wrong places. Rot is the pipelines running for reports nobody reads any more. None of it is exotic. All of it compounds. And once it compounds, no AI model can outrun it.

Build for the human AND the bot

One of the most useful frames in the report is what WPP calls the “two rails” — the human rail (experiences, trust, reviews, editorial, brand signals) and the bot rail (structured product data, taxonomy, schema, technical infrastructure that makes the business discoverable and recommendable to AI systems).

The bot rail sits on the same data foundation as internal analytics. The same product master, the same taxonomy, the same schema. Which means the “fix your data warehouse” argument isn’t just about internal efficiency any more — it’s about whether an AI agent recommending a golf driver to a customer can even find, parse and evaluate your product data at all. In an agent-mediated shelf, being unreadable is being invisible.

“The agent’s shelf is radically smaller. Being accurately represented, interpreted and considered by AI systems matters even before agent-led purchasing becomes mainstream.”

The report cites a case where WPP rewrote 77 product pages for a consumer electronics brand on Amazon so the marketplace’s AI shopping assistant could read them properly. No change to photography, pricing or media spend. Sales up 84%. Conversion up 87%. That’s not a marketing story — it’s a data-readability story.

What retail CDOs should actually do

The report closes with six recommendations. In our experience the ones that unlock the rest are these three:

• Build the roadmap before scaling spend. Define the business problem, the ownership, the governance and the success measure before fragmented experimentation becomes fragmented investment. This is what our Discovery Call + Diagnostic sequence is designed to force.
• Treat data and architecture as growth infrastructure. Strengthen product data, integration, taxonomy, governance and machine readability alongside the AI use cases. Not before. Not after. Alongside. Both rails need to be laid at the same time — but one has to sit under the other.
• Measure what the agent sees. Your existing dashboards can’t show you agentic performance. Traffic and revenue can look fine while an agent quietly routes the consideration set around you. If AI-mediated discovery is even 10% of your journey today, you need visibility into it now — because it’s only going to become more.

The fix isn’t a bigger AI budget

The most consequential line in the whole report, from Section 1: “The risk is not moving too slowly. It is spending faster than the organisation can absorb, govern and measure.”

For every retailer we’ve worked with that stalled on an AI initiative, that sentence was the postmortem in advance. They approved another AI project when what they needed was to audit the data platform that would have to feed it. A Snowflake environment audit — five days, fixed fee, board-ready PDF — surfaces the four axes (cost, security, freshness, data usefulness) that determine whether an AI initiative is even shippable. It’s the cheapest, fastest way to close the readiness gap the report is pointing at.

Fix the warehouse first. Then build the AI on top. The report just made that a lot easier to justify in the boardroom.